Efficient coordination of several robot arms in order to carry out some given independent/cooperative tasks in a common workspace, avoiding collisions, is an appealing research problem that has been studied in different robotic fields, with industrial and service applications. Coordination of several robot arms in a shared environment is challenging because complexity of collision free path planning increases with the number of robots sharing the same workspace. Although research in different aspects of this problem such as task planning, motion planning and robot control has made great progress, the integration of these components is not well studied in the literature.

This thesis focuses on integrating task and motion planning multi-robot-arm systems by introducing a practical and optimal interface layer for such systems. For a given set of speci fications and a sequence of tasks for a multi-arm system, the studied system design aims to automatically construct the necessary waypoints, the sequence of arms to be operated, and the algorithms required for the robots to reliably execute manipulation tasks.

The contributions of the thesis are three-fold. First, an algorithm is introduced to integrate task and motion planning layers in order to achieve optimal and collision free task execution. Representation via shared space graph (SSG) is introduced to check whether two arms share certain parts of the workspace and to quantify cooperation of such arm pairs, which is essential in selection of arm sequence and scheduling of each arm in the sequence to perform a task or a sub-task. The introduced algorithm allows robots to autonomously reason about a structured environment, performs the sequence planning of robots to operate, and provides robots and objects path for each task to succeed a set of goals.

Secondly, an integrated motion and task planning methodology is introduced for systems of multiple mobile and fixed base robot arms performing different tasks simultaneously in a shared workspace. We introduce concept of dynamic shared space graph (D-SSG) to continuously check whether two arms sharing certain parts of the workspace at different time steps and quantify cooperation of such arm pairs, which is essential to the selection of arm sequences and scheduling of each arm in the sequence to perform a task or a sub-task. The introduced algorithm allows robots to autonomously reason about complex human involving environments to plan the high level decisions (sequence planning) of robots to operate and calculates robots and objects path for each task to succeed a set of goals.

The third contribution is design of an integration algorithm between low-level motion planning and high-level symbolic task planning layers to produce alternate plans in case of kinematic and geometric changes in the environment to prevent failure in the high-level task plan.

In order to verify the methodological contributions of the thesis with a solid implementation basis, some implementations and tests are presented in the open-source robotics planning environments ROS, Moveit and Gazebo. Detailed analysis of these implementations and test results are provided as well.